New MacBook Pros With M2 Pro and M2 Max Chips Now 'Well Into Development'

[theluggage]

I'm not convinced we'll get an annual upgrade cycle. (We didn't on the MacBook Air.)

Or the Mac Mini. Or the 24" iMac.

Maybe we'll get a MBP update by the end of the year, but I don't think anybody outside forums like this is going to be sending out a search party until late next spring.

Also, there's now no direct like-for-like CPU/GPU comparison between Mac and PC, Apple don't need to worry about competing with each incremental improvement Intel makes - e.g. Dell trumpeting 14th generation Intel Core while the MacBook Pro is still on 12th gen. They've got a power/performance benefit over Intel that isn't going away in one generation - and are probably going to beat Intel to 3nm which should replenish that lead in a year or so. I don't see a huge urgency to upgrade the MBPs with some stop-gap 5nm tech.

(Also, some of the M2's improvement over the base M1 - e.g. LPDDR5 - already went into the M1 Pro/Max, so M1 Pro to M2 Pro might not exactly be night and day...)

You have two nice choices: either buy a relatively new (i.e., cutting edge) product, since it just came out a month ago; or wait six months for the next gen.

Vs. buy an 8-month old model which has had all the teething troubles fixed and has proven not to be a lemon - and which won't force you to upgrade to the latest OS before it's got the the 3rd or 4th update. Then be able to get another one of the same model 6 months later when you add another person to the team.

These days, I think buying a new machine every 3-5 years is sensible, and an 18-month upgrade cycle is short enough to ensure that you're not going to get forced into buying an outdated machine. For the people who must have this year's machine then Apple could always introduce new colours every season...

Problem is, machines like the 2014 Mini and the trashcan Mac Pro have previously been on a 4-6 year update cycle - not so good - and that wasn't all on Intel.

 

[theluggage]

Perhaps. Or perhaps the M1 Max was developed in two forms: one with the single UltraFusion connector for Max and Ultra and then another with two UltraFusion connectors for the "Extreme" which was only going into the Mac Pro.

Where would the second connector go? From the photos of the M1 Pro and Max it looks like the two long sides are occupied by the RAM connections (plus, that's where the RAM physically fits on the package - where would that go if you had two pairs of Max's side-by-side?) and one of the short sides holds the Thunderbolt and other ports.

Personally, I was assuming some sort of long, thin "hub" chiplet with two Ultrafusion connectors on each long edge. Still leaves the question of where the RAM chips would sit.

I guess someday we'll find out what the true plan was, though it's academic at this point since the model never left the lab.

Maybe it never left software simulation - and the second connector was located in hyperspace. Easy to program. Harder to build

Seriously though - you probably could simulate a 4x M1 Max system "logically" to get an idea of how it would perform before worrying about how you'd actually make it.

Anyway, I don't think anybody has seen a picture of what the new M? Max chip is going to look like - it could be radically different.

Here's a random thought (no basis): will newer (probably 3nm - which could allow higher core counts) chips even follow the same vanilla/Pro/Max/Ultra cadence? With the M1 series, the two fundamental die designs are the M1 and M1 Max - the M1 Pro is (design wise) just the top 2/3 of a Max die and the Ultra is two Max's joined together.

What if a future Mn series did it differently? E.g the "Pro" could be two regular Mns joined together, the "Ultra" could be a monolithic 20 core monster, the "Max" the top 2/3 of an Ultra and the Extreme/Ludicrous/Quadra/whatever for the Mac Pro could be two Ultra's joined together... Invent your own permutations and names.

 

[CWallace]

Where would the second connector go? From the photos of the M1 Pro and Max it looks like the two long sides are occupied by the RAM connections (plus, that's where the RAM physically fits on the package - where would that go if you had two pairs of Max's side-by-side?) and one of the short sides holds the Thunderbolt and other ports.

Again, all I have is the reports from that Yujin(sic) fellow that claims Apple designed the UltraFusion interposer to work both as it does on the Ultra as well as on the "sides".

If it can't work with the Max as we know it, then it's possible the "M1 Extreme" was a custom Max-based SoC with a different physical layout. One assumes this SoC would also have things the Max does not like external memory controllers (to address off-package DIMM RAM) and the ability to drive external PCIe expansion slots so a custom layout does make some sense.

Personally, I was assuming some sort of long, thin "hub" chiplet with two Ultrafusion connectors on each long edge. Still leaves the question of where the RAM chips would sit.

There has been no shortage of mock-up layouts using PhotoShop's Clone tool and the M1 Max and M1 Ultra by YouTubers, to be sure.

Maybe it never left software simulation - and the second connector was located in hyperspace. Easy to program. Harder to build

Seriously though - you probably could simulate a 4x M1 Max system "logically" to get an idea of how it would perform before worrying about how you'd actually make it.

Indeed. I would not at all be surprised if the "M1 Mac Pro" was effectively a prototyping board to test how the thing worked and bore zero resemblance to what Apple would have eventually shipped if the model had gone to market.

I remember my days at Microsoft when they were testing the original XBOX in the cafeterias and it was literally a collection of parts bolted to a plywood board that they rolled around and connected to the TVs.

Anyway, I don't think anybody has seen a picture of what the new M? Max chip is going to look like - it could be radically different.

True. There have been absolutely zero leaks about the physical design of the Mac Pro class SoC and now that the M1 model never made it out of the labs, we arguably no longer even know the CPU/GPU/NE core counts.

 

[chucker23n1]

If it can't work with the Max as we know it, then it's possible the "M1 Extreme" was a custom Max-based SoC with a different physical layout. One assumes this SoC would also have things the Max does not like external memory controllers (to address off-package DIMM RAM) and the ability to drive external PCIe expansion slots so a custom layout does make some sense.

Maybe, but that's quite different from the original Jade 4C-Die rumor, which suggested four M1 Max chips, presumably in a 2x2 grid.

 

[theorist9]

Vs. buy an 8-month old model which has had all the teething troubles fixed and has proven not to be a lemon - and which won't force you to upgrade to the latest OS before it's got the the 3rd or 4th update. Then be able to get another one of the same model 6 months later when you add another person to the team.

These days, I think buying a new machine every 3-5 years is sensible, and an 18-month upgrade cycle is short enough to ensure that you're not going to get forced into buying an outdated machine. For the people who must have this year's machine then Apple could always introduce new colours every season...

Problem is, machines like the 2014 Mini and the trashcan Mac Pro have previously been on a 4-6 year update cycle - not so good - and that wasn't all on Intel.

Just to be clear, I wasn't talking about an annual redesign, just an annual spec bump to the latest microarchitecture and (if available) latest process tech. An annual Mac uarch update makes sense to me, since the Mac uarch is based on the iPhone uarch, which does update annually, so that would enable them to synchronize the two.

The comment about the colors is unnecessary snark, since those who would like to see more frequent spec bumps, for whenever they are in the market, are likely those who prize function over form. And also to be clear: I wasn't arguing for an annual spec bump for people who want to update their Macs annually (surely a tiny segment); instead, it's for the benefit of those who upgrade more normally (every 3- 6 years), enabling them to get the latest tech whenever they're in the market.

And this is personal preference, but I don't have as much concern about teething troubles for spec bumps as you do.

Finally, and conversely, if there's somewhere I think Apple might want to do longer update cycles, because the teething troubles with each new release are extensive, it's to the OS. [Unless, they go with annual uarch updates and, for reasons of synergy, need to do annual OS updates as well—so maybe if they instead do go on an 18 month uarch update cycle for the Mac, they should also do an 18-month OS cycle?]

 

[chucker23n1]

Just to be clear, I wasn't talking about an annual redesign, just an annual spec bump to the latest microarchitecture and (if available) latest process tech. An annual Mac uarch update makes sense to me, since the Mac uarch is based on the iPhone uarch, which does update annually, so that would enable them to synchronize the two.

That the µarch exists for iPhone doesn't mean that the SoC design needs to exist for Mac.

 

[theorist9]

That the µarch exists for iPhone doesn't mean that the SoC design needs to exist for Mac.

True, it's not a requirement, but I never said it was. And obviously they're not doing that now. Still, it seems it would be beneficial if two were synchronized, since that would give us annual Mac uarch updates, and it might also enable a synergy between the iPhone and Mac chip design teams.

 

[chucker23n1]

it seems it would be beneficial if two were synchronized, since that would give us annual Mac uarch updates, and it might also enable a synergy between the iPhone and Mac chip design teams.

I think that’s what Craig and Sruji would like, yes. Having the entire software and hardware stack be in sync would be very nice for engineers.

But I also suspect (hope?) they’re starting to see it’s too hard, and that they need to uncouple some portions.

 

[Tagbert]

I think that’s what Craig and Sruji would like, yes. Having the entire software and hardware stack be in sync would be very nice for engineers.

But I also suspect (hope?) they’re starting to see it’s too hard, and that they need to uncouple some portions.

I don't think it is feasible to not do software releases but may be better to decouple basic device support from new features and from application updates so that things aren't forced to ship just because the hardware is coming. And to devote just a little more dev resources to bugfixes.

 

[name99]

Again, all I have is the reports from that Yujin(sic) fellow that claims Apple designed the UltraFusion interposer to work both as it does on the Ultra as well as on the "sides".

If it can't work with the Max as we know it, then it's possible the "M1 Extreme" was a custom Max-based SoC with a different physical layout. One assumes this SoC would also have things the Max does not like external memory controllers (to address off-package DIMM RAM) and the ability to drive external PCIe expansion slots so a custom layout does make some sense.




There has been no shortage of mock-up layouts using PhotoShop's Clone tool and the M1 Max and M1 Ultra by YouTubers, to be sure.




Indeed. I would not at all be surprised if the "M1 Mac Pro" was effectively a prototyping board to test how the thing worked and bore zero resemblance to what Apple would have eventually shipped if the model had gone to market.

I remember my days at Microsoft when they were testing the original XBOX in the cafeterias and it was literally a collection of parts bolted to a plywood board that they rolled around and connected to the TVs.




True. There have been absolutely zero leaks about the physical design of the Mac Pro class SoC and now that the M1 model never made it out of the labs, we arguably no longer even know the CPU/GPU/NE core counts.

Ultrafusion is the protocol, it does not depend on particular hardware; specifically it is not tied to an interposer or a bridge chip. The M1 Ultra uses a bridge chip (like Intel's much ballyhoo'd EMIB, though much more performant). But that corresponds to a 2014 patent. Two patents (2017 and 2018) describe an alternative way to "glue" the chips together that does not require this bridge chip (so presumably a little cheaper, and a little lower power). Who knows when it will be till those later patents become products?

The second point is about the geometry. The current M1 designs direct attach DRAM to two sides of the chip, which means if you want to join multiple chips together, the best you can do is create a lone line of them, say three or four in a row. A 2018 patent describes multiple ways Apple envisions (probably likely to play out over the next decade) for getting around this. The patents envision
(a) the base unit of interest is something like an Ultra, two dies directs attached
(b) explicit "bridges" join these two-die units together in various ways (this is very much something like current AMD designs with an Ultra corresponding to an AMD chiplet, and the bridges corresponding to Infinity Fabric links(
(c) "memory spikes", long slivers of silicon plus wiring have DRAM attached to them, and are attacked to particular anchor points on the chips. A picture makes all of this more obvious:
(I don't know why this PoS interface won't allow me to insert a picture, but hopefully this attached image will work.)

OK, finally. Freaking amateur hour on this website! Won't accept PNG or PDF images, only JPG. Enjoy all its blurry goodness!

 

[Carlo Aguilar Castrat]

Well, I still hold out hope for the October event or thereabouts. The earlier predictions were for an M2 Mac mini alongside the M2 MacBook Air, and then for an M2 Mac mini release in the fall.

As for M2 Pro, who knows if that will even be released.

I kind of hope it won't be, I just bought an M1 pro 14 inch today as a $250 discount and I for some reason I thought this one came out not too long ago, now I'm worried the m2 pro will come out next month and I made a bad choice buying the m1 pro one now

 

[CWallace]

Kuo today said the refreshes for the 14" and 16" will likely only replace M1 family SoCs with M2 family SoCs so unless you have a workflow that would benefit from that, saving the $250 might be the better value proposition.

 

[richinaus]

Kuo today said the refreshes for the 14" and 16" will likely only replace M1 family SoCs with M2 family SoCs so unless you have a workflow that would benefit from that, saving the $250 might be the better value proposition.

I agree.

The M1 in the 14 is great - I wouldn't be overly concerned about the M2 unless I actually needed something that it offers [GPU / encoding etc].

Personally, I never had any expectation that the M2 MBP would be any different to the M1 MBP except for the chip upgrade.

 

[Carlo Aguilar Castrat]

Kuo today said the refreshes for the 14" and 16" will likely only replace M1 family SoCs with M2 family SoCs so unless you have a workflow that would benefit from that, saving the $250 might be the better value proposition.

I actually just realised it will be about 500 dollars saved instead of 250, so now I'm fine with it. The iPhone 14 prices from Apple in Europe are way higher than they were last year (30% higher in euros than last year) (because the euro has devalued). So I just bought this M1 Pro 14 inch for $1992 on discount (tax included) with 3 years warranty included (3 years minimum required by law here now) and Based on the price increases to the iPhones, if the new 14 inch is the same price in the US, 25-30% more expensive here, so it will be about 500-600 dollars more than what I just spent on the m1 pro 14 inch. A

 

[ScooterComputer]

The M1 in the 14 is great - I wouldn't be overly concerned about the M2 unless I actually needed something that it offers [GPU / encoding etc].

This… I've been telling clients asking about the M2 Air to get the $1599 M1 Pro, uh, Pro at BB… I think it is a significantly better value. It outperforms the M2, plus and the screen is gorgeous.
The only thing I see coming in the M2 Pro that will be "significant" (beyond baseline performance improvement, 15-20%) is with the Media Engine, the enhanced/expanded encoders, and that certainly won't be something that users outside of video production really will need in the short term. (AKA, folks the M2 Air is targeted at: who surf the web, email and message, run Office). Coincidentally, those are also the tasks—heavy encoding video jobs—that seem to hammer the single-channel SSD in the $1200 M2 Air too, which force-pushes those users up into the $1400/$1500 unit pricing anyhow.
IMHO, the strategy Apple chose—keeping the M1 Air ("pricing") while hobbling the M2 Air (diminishing "value") really is disappointing. Worse, is it's likely the community will have to wait for another 18 months for it to get fixed (and, at that point, folks will be purchasing a $1000 3 year old computer).

 

[kp98077]

Are people still thinking the M2 MBP will be out late 2022?

 

[CWallace]

Are people still thinking the M2 MBP will be out late 2022?

If Kuo is correct and all that will change is M1 Pro/Max will be swapped out for M2 Pro/Max, then I think it could launch in Oct/Nov at an Apple Event.

 

[kp98077]

If Kuo is correct and all that will change is M1 Pro/Max will be swapped out for M2 Pro/Max, then I think it could launch in Oct/Nov at an Apple Event.

okay thanks, assume he is usually correct? lol..

 

[benspratling]

Mac mini refreshment cycle went again back to 4 years.
So first "new" Mac mini in 2024 wit M3 chip (no Pro/Max).

I was thinking the same thing. It could be the only reason Mac Minis got M1's before MacBook pros was Apple didn't have the faster chips yet. With the Mac Studio in the picture, and a Mac Pro yet to be transitioned to M# chips, Mac mini may linger precisely so they will remain a low-cost option, while Apple spends its efforts on bringing the higher-end chips to market.
On the other hand, they do like to rev the chips in their iPhones and iPads every year, so maybe the plan is just rev everything every year. But there may be a catch-up period while higher-end chips are created. AppleTV's use lower-end chips than phones and tablets, I assume that's at least partially to keep costs down.

 

[Allen_Wentz]

I was thinking the same thing. It could be the only reason Mac Minis got M1's before MacBook pros was Apple didn't have the faster chips yet. With the Mac Studio in the picture, and a Mac Pro yet to be transitioned to M# chips, Mac mini may linger precisely so they will remain a low-cost option, while Apple spends its efforts on bringing the higher-end chips to market.
On the other hand, they do like to rev the chips in their iPhones and iPads every year, so maybe the plan is just rev everything every year. But there may be a catch-up period while higher-end chips are created. AppleTV's use lower-end chips than phones and tablets, I assume that's at least partially to keep costs down.

I think not. IMO Apple needs to fill the space between the $700/$900 M1 Minis and the $2k Studio ASAP because the $1100 Intel Mini needs to go away. A stronger M2 Mini could fill that $1100-$1400 price point, perhaps without all the features of the Intel Mini.

Edit: Needs to go away may be too strong. Apple should probably leave the Intel Mini unchanged but still available for those entities that must have an Intel box for some reason.

 

[benspratling]

If they do release an M2 MBP this year I'd seriously consider a "midnight" one... (I mean, assuming they offer that colour, obviously.) I don't love being on the absolute first iteration of the new hardware, since historically these don't tend to be the greatest machines to own long term. But we'll see... The only problem I have with my current machine (16" M1 Pro MBP) is that I get occasional audio dropouts... But I've seen others running Monterey on Intel machines mentioning this as well, so I'm not super confident that the hardware is to blame.

(In fact, it's often correlated with switching apps, or other operations that involve a significant change to the window content, which makes me think some jackass on the macOS dev team did something really stupid in a bit of drawing code that's allowing it to mess with the audio thread...)

Weird, I get airdrop... dropouts. I have to restart my m1 max machine for it to work again. I've seen audio "skipping" on intel macs for years, ever since they introduced new "improved" audio buffering algorithms into the os. It's at the point where I'm having a hard time imagining someone using a mac to play back a video at a meeting. We had more reliable and smoother playback in 2005!

 

[benspratling]

I think not. IMO Apple needs to fill the space between the $700/$900 M1 Minis and the $2k Studio ASAP because the $1100 Intel Mini needs to go away. A stronger M2 Mini could fill that $1100-$1400 price point, perhaps without all the features of the Intel Mini.

You can configure an m1 mac mini all the way up to $1800.

 

[opeter]

I think not. IMO Apple needs to fill the space between the $700/$900 M1 Minis and the $2k Studio ASAP because the $1100 Intel Mini needs to go away. A stronger M2 Mini could fill that $1100-$1400 price point, perhaps without all the features of the Intel Mini.

Edit: Needs to go away may be too strong. Apple should probably leave the Intel Mini unchanged but still available for those entities that must have an Intel box for some reason.

It will go away. As did the iMac Pro and the iMac 27" go away... simply removed from the store.

 

[Allen_Wentz]

You can configure an m1 mac mini all the way up to $1800.

Of course. One can also configure a Studio up to $8k, but that is not the point. The point is price lining. There is generally a space above the M1 Mini and below the Studio that is currently occupied by the Intel Mini. Since Apple is unlikely to want to continue improving the Intel Mini (I could be wrong; Apple may want/need to solidly support legacy Intel for edu or for corporate) it seems to me that there should be minimum two flavors of M-series Mini. One lowest end (perhaps they keep the M1 Mini for that role) and a stronger M2 Mini.

Personally I believe that with the new (excellent) Studio now in place the Minis should clearly be positioned as entry level boxes. I am aware that many folks want to configure Minis as less expensive Studios, but that does not seem like wise product placement for Apple. Just my $0.02.

 

[Allen_Wentz]

Weird, I get airdrop... dropouts. I have to restart my m1 max machine for it to work again. I've seen audio "skipping" on intel macs for years, ever since they introduced new "improved" audio buffering algorithms into the os. It's at the point where I'm having a hard time imagining someone using a mac to play back a video at a meeting. We had more reliable and smoother playback in 2005!

That is the kind of thing that is the reason I prefer to wait for v2. Hopefully M2 will correct that (even if it is at its heart a software issue).